Private credit deal screening, explained.

Sponsors, bankers, teasers and CIMs: how a buyout loan reaches a direct lender's desk

The life of one deal at a direct lender, in five steps

1. The banker sends a teaserFour pages, no company name. The fund signed Lantern's confidentiality agreement on Monday, October 5, 2026, and logged the teaser on Tuesday, October 6.
2. The data room opensThe CIM, the quality of earnings report, the model and the process letter: 192 pages in all. The IOI is due Friday at 5 p.m. Eastern.
3. The underwriter screens itMandate fit, the fund's own EBITDA, leverage and cover, the sponsor, the first red flags. This is where Lantern's 6.08x is found, or the 5.26x is believed.
4. Pursue or pass, then an IOIThe head of underwriting approves spending diligence money; the banker gets a range, terms and first questions.
5. Diligence, committee, a commitmentWeeks of meetings and lawyers. A deal that fails a limit here, after the money is spent, is the cost a good screen prevents.

Step 3 is the subject of this guide. The fund, the companies, the people and the figures in this example are fictional.

The seller's EBITDA, quiet defaults and the leverage guidance that is gone

Add-back, leverage, EBITDA and unitranche

Fourteen terms are defined in the one-page summary.

Teaser, CIM, quality of earnings and process letter: what a direct lender reads before an indication of interest

In shortA deal reaches a fund as a stack of documents written to sell it. Each answers a different question, and each is written by someone with an interest: the banker writes the teaser and the CIM, the sponsor's accountants write the quality of earnings report, the sponsor builds the model. The screen reads them against each other and against the fund's own rules.

What a teaser can settle, it should. A fund that does not lend to restaurants, oil and gas exploration or cannabis can pass on the teaser and save the CIM for deals it can make. The CIM and QoE are where the hours go, and where the screen earns its keep: the CIM presents the seller's view, the QoE shows which of the seller's adjustments the accountants were willing to support, and the differences between the two are the first questions to ask.

One vendor blog cites a private equity origination benchmark, Sutton Place Strategies' 2024 report, for 80 to 100 memoranda reviewed for each deal closed; the same article also words it as 80 to 100 a year.7 That figure describes buyout firms looking for companies to buy, and we found no published equivalent for lenders. The number a lending team needs is its own: memoranda received, read in full, pursued and closed.

The rules behind a direct lender's first screen, and what changed in 2025 and 2026

In shortNo regulation tells a fund how to screen a deal. The fund's own mandate does, inside a few hard rules: a BDC's 70% test on what it buys, its borrowing limit, the tax diversification test most BDCs must pass, and an adviser's duty to keep records. Bank lenders had leverage guidance until the OCC and FDIC withdrew it in December 2025. What changed most is attention: the SEC, the Fed and the Financial Stability Board all looked hard at private credit in 2026.

Mandate fit and concentration limits: testing the loan, then the whole fund after the hold

In shortA mandate has two kinds of test. Credit tests ask whether this loan is one the fund should make: sector, size, leverage, add-backs, interest cover (EBITDA divided by interest), sponsor equity, top customer, price. Portfolio tests ask what the fund looks like after its share of the loan, its hold, is added: its share in one borrower, one sector and one sponsor, and its qualifying assets. A loan can pass every credit test and still break a portfolio limit.

The twelve tests in Lantern's screen

The arithmetic of a portfolio test

A portfolio test adds the proposed hold to what the fund already holds and divides by the fund after the loan. Fund IV has $2.74 billion invested and $318 million of it in industrial services, 11.6%. With a $90 million hold in Lantern, industrial services becomes $408 million of $2.83 billion: 14.4%, under the 15% cap, with $16.5 million of room left in the sector.

The next industrial services deal of that size would not fit. A screen that reports only "pass" hides that; one that reports the room left tells origination where the fund can still lend.

Sponsor limits work the same way. A fund with several loans to one sponsor's companies depends on how that sponsor behaves in each of them. Lantern's sponsor already has $212 million with the fund across six loans; adding $90 million takes it to $302 million, 9.4% of the fund's $3.2 billion of commitments, under a 12% cap. The single-borrower limit is the simplest test and the one bankers ask about first: lenders were invited to hold $75–100 million, and the fund's cap is $112 million.

Federal tax rules put an outer bound on the same idea for funds taxed as regulated investment companies: no more than 25% of assets in one issuer, measured each quarter.18 Most mandates are far tighter. Direct lenders are also more specialized by industry than banks, according to recent research on the market,22 so a sector cap deserves as much care at the screen as a single-name cap.

EBITDA add-backs: rebuilding the sponsor's adjusted EBITDA under the fund's own policy

In shortStart from reported EBITDA, built from the financial statements. Then take each add-back the seller proposes and decide, under a written policy, whether it was incurred, whether it will recur, and whether there is evidence. One-time costs that were paid usually pass. Savings not yet achieved and price increases not yet in the numbers usually do not. The studies agree that generous add-backs predict worse outcomes.

Lantern's bridge, add-back by add-back

Three tests for every add-back

Projected adjusted EBITDA ran about 30% above what deals earned

Three bodies of work point the same way. S&P Global Ratings has published a series of studies on add-backs in buyouts and mergers since 2018. Its November 24, 2020 edition found that projected adjusted EBITDA at the start of a deal exceeded the EBITDA actually earned in the following two years by about 30% on average, as summarized by Federal Reserve Bank of St. Louis researchers.2,3 For deals made in 2015, S&P reported add-backs equal to 50% of the EBITDA that B-rated companies estimated, and 21% for BB-rated companies.23

A Federal Reserve Bank of St. Louis working paper by Faria-e-Castro, Gopalan, Pal, Sánchez and Yerramilli (September 2022) read the EBITDA definitions in 3,939 syndicated loan agreements from 1995 to 2019. All but 344 allowed at least one add-back that departs from accounting rules, and about 43% allowed three or more.

To measure performance, the authors used a smaller set of 411 loans matched to supervisors' Shared National Credit data and made between May 2011 and May 2015: each additional add-back raised the chance that a loan was 60 days delinquent within three years by 4.2 percentage points (against a 1.3% base rate), and the chance of default within three years by 1.1 points, as much again as the 1.1% base rate.3

The third body is the documents themselves. In law firm Proskauer's own 2025 private credit deals, add-backs for run-rate cost savings and synergies were "more or less ubiquitous", and only 49% of deals with less than $50 million of EBITDA capped add-backs for non-recurring expenses, down from 67% in 2023. Among larger deals, 26% had such a cap.24

Cash tells the same story from another side. KBRA, a rating agency that assesses about 2,400 middle market sponsored borrowers, reported in May 2026 that their median EBITDA grew at 27% a year over two years while operating cash flow grew 8%; cash flow fell from 33% of EBITDA to 21%.25 An add-back adds to EBITDA. It never adds cash.

Leverage and interest cover at entry: the two ratios that size a unitranche

In shortLeverage is debt divided by EBITDA; interest cover is EBITDA divided by cash interest. Both use the fund's EBITDA, and cover uses today's floating rate. The largest loan a mandate allows is the leverage limit times the fund's EBITDA. Private credit borrowers carry more debt and less cover than syndicated loan borrowers, according to the Federal Reserve.

The arithmetic on Lantern

Cover is where floating rates bite. Unitranche loans pay a margin over SOFR, the overnight rate set in the market for Treasury repurchase agreements. SOFR rose from 3.62% on September 16, 2026 to 3.85% the next day and stood at 3.94% on October 29.26 A screen that tests cover at last quarter's rate will pass loans that fail at today's.

How Lantern compares with the market

The Federal Reserve compared loans to private credit borrowers with leveraged loans in the syndicated market in an August 2026 note. The median private credit borrower had debt of 5.0 times EBITDA, against 3.2 times for leveraged loan borrowers, and average cover of 2.2 times cash interest, against 3.7 times.

Private credit loans were smaller (a median of $20 million) and paid a higher spread (500 against 400 basis points over the base rate, that is 5.00 against 4.00 percentage points), and 85% of the borrowers were rated or estimated single-B, with 15% at CCC or below.27

Other measures put middle market leverage higher. KBRA's median total leverage across the borrowers it assessed in June to August 2026 was 6.5 times,28 and its median interest cover in the first quarter of 2026 was 1.6 times.29 In Proskauer's own 2025 deals, closing leverage averaged 5.1 times, up from 4.6 times in 2023, with 68% of deals between 4.0 and 7.0 times.24 Each source measures EBITDA its own way, which is one more reason a fund should compare deals on its own definition.

Reading the sponsor: its record with the fund, its equity cushion, and how it behaved under stress

In shortIn sponsor-backed lending, the owner matters almost as much as the company. A sponsor that puts in a large share of equity, a history of putting more money in when a company stumbles, and a record of honest reporting is worth a lot to a lender. Research supports the habit: private equity owners tend to resolve distress faster and out of court. The best evidence is the fund's own history with that sponsor.

Survey evidence shows how much lenders lean on sponsors. In a survey of private debt funds managing over $300 billion, Block, Jang, Kaplan and Schulze found that funds say private equity sponsors help them lend more and write more effective covenants, the financial tests a borrower must keep passing.30

Research on outcomes supports that view. Hotchkiss, Smith and Strömberg studied 2,151 leveraged loan borrowers from 1997 to 2010: firms owned by private equity were no more likely to default than other borrowers with the same leverage, and when they did default they restructured more often out of court and faster, and sponsors that put in new money before distress saw better restructurings.31

Those are averages. What a screen needs is this sponsor's record. For Lantern, the fund's files show six earlier loans to companies owned by the same sponsor since 2019: four active, two repaid at gross returns of 11.2% and 12.6%, no defaults, one covenant reset in 2024 at a dental group where the sponsor put in $18 million of new equity. Average leverage at closing was 5.13 times. At 6.08x, Lantern as asked would carry more debt for its earnings than any earlier loan to this sponsor's companies.

Term sheet red flags a first screen should catch before the lawyers start

In shortAt the IOI stage the fund has only a process letter, maybe a draft term sheet, and the CIM's financing page. Even so, a few terms change what the loan is worth: no maintenance covenant, an EBITDA definition with uncapped add-backs, loose room to borrow more, PIK options, and generous cures. The screen should list them as questions now, while the fund's answer can still shape the terms.

Lantern's financing page shows a $285 million unitranche, a $35 million revolving credit facility and a $40 million delayed-draw term loan, priced at SOFR plus 5.00–5.25% with an original issue discount of 98.5 (the fund advances $98.50 for each $100 of loan). The screen's leverage uses the $285 million funded at closing. If the delayed-draw loan is used in full, debt becomes $325 million, 6.93 times the fund's EBITDA before any acquired earnings arrive. The first questions should ask what test each draw must meet.

Ten ways a first screen goes wrong, and the check that stops each one

In shortMost screening errors flatter the deal: the seller's EBITDA taken as given, cover tested at an old rate, a portfolio limit checked after the diligence money is spent, a pass never written down. Each has a simple check. The hard part is running every check on every memo, including the ones read at 9 p.m.

Private credit in 2026: about $1.4 trillion of loans and four different default rates

In shortPrivate credit has roughly doubled in five years. The Federal Reserve puts it at about $1.4 trillion of US loans; the Financial Stability Board puts global assets at $1.5–2.0 trillion. Default rates in 2026 range from about 2% to 6.3% depending on who counts and what counts as a default. Recoveries on direct loans have been lower than on syndicated loans.

The growth came with links to the rest of the system. FSOC reported in December 2025 that banks' loan commitments to private credit funds were about $445 billion in the second quarter of 2025,32 and the Fed counted $2.6 trillion of bank commitments to nonbank financial institutions at the end of 2025, with private equity, BDCs and private credit the largest category.1

About 20% of private credit fund assets now sit in vehicles that let individual investors redeem quarterly; in the first quarter of 2026, accepted redemptions at perpetual BDCs exceeded new money for the first time, and most managers capped redemptions at 5% of net assets.1

Four default rates, each right for its own scope

Fitch's rate is the highest because it counts what most private credit distress looks like. It has risen from 5.8% for the twelve months to January 202636 to 6.3% in August; in the year to August, 89 issuers defaulted, and smaller borrowers fared worst: 12.0% for issuers with EBITDA under $25 million, 5.2% for $26–50 million. Healthcare providers and industrial and manufacturing companies were both at 9.9%; software, the largest sector in private credit portfolios, was at 0.6%.33,4,1

Proskauer's index tracked 746 loans with $205.7 billion of original principal and fell to 2.41% in the third quarter of 2026 from 2.73% in the first.34,37

When loans do default, recoveries are lower than in syndicated loans

The Federal Reserve's February 2024 note cites these figures from KBRA DLD without a date range; treat them as a snapshot. A lower recovery means a first screen's leverage and equity cushion matter more, because there is less to fall back on.38

For the screen, the useful facts are the differences. Small borrowers default more in Fitch's data and in Proskauer's ($25–49.9 million EBITDA at 3.3% in the third quarter, against 1.7% at $50 million and above).34 New loans pay less: BDC origination spreads ran 55–60 basis points below their existing portfolio spreads in early 2026, and 77% of BDCs marked their term loans down in the first quarter.35 A loan priced at the bottom of the range, to a smaller borrower, needs a cleaner screen than the market average suggests.

Supervisors, the SEC and private credit: what changed between 2025 and October 2026

In shortBank supervisors relaxed one rule and sharpened their questions. The OCC and FDIC withdrew the leveraged lending guidance; the Fed asked the largest banks about their private credit exposure and launched a survey of direct lending. The SEC named private credit an exam focus, its staff set out expectations on valuing private loans, and retail money is arriving through 401(k) plans and BDCs. Each of these raises the value of a written, dated record of why a fund made or passed on a loan.

Together these changes widen who may ask to see a screen: an examiner, a bank lending to the fund, an auditor testing a valuation, a retail fund's board. A screen that names its sources, its definitions and who decided is easier to show than one that lives in an analyst's spreadsheet.

AI in first-pass deal screening: what software can read, and what stays with the underwriter

In shortSoftware can read every page of every memorandum on the day it arrives, key every figure with its page, rebuild EBITDA under a written policy, run every mandate and portfolio test, and draft the questions. Judging management, pricing risk and deciding remain with people. No SEC rule addresses advisers' AI use specifically since June 2025; examiners check that what firms say about their AI is true, and banks follow the April 2026 model risk guidance.

Who keys the CIM, who rebuilds EBITDA and who spends the diligence budget at a direct lender today

In shortAn associate keys the memorandum into a screening template; a vice president rebuilds EBITDA, runs the tests and writes the screen; a head of underwriting decides whether to spend diligence money; the investment committee approves large indications. Most of the hours go to reading and keying. Teams screen in house, with outsourced analysts, or with software, and most use a mix.

Where the hours go on one memorandum

Associates keying CIMs, an offshore team or deal screening software

Project Lantern's 192-page memorandum, marked up against the mandate, ending in a smaller loan approved to pursue

In shortThis is the example from the top of the page, inside the product. Six agents read Lantern's 192 pages in 6 minutes 12 seconds after the data room opens. Daniel Okafor, the underwriting vice president, opens the memorandum already marked up against the fund's mandate, sees leverage on the fund's own EBITDA, and recommends a smaller loan with six questions. Elena Marsh, the head of underwriting, approves pursuing it.

1

His day opens on the memorandum, already marked up

Daniel's day starts in the Lantern CIM itself, which the agents read between 7:58 and 8:04 a.m. Every sentence a mandate test depends on is highlighted: green where it fits, red where it breaks a limit, amber where it needs watching. Each carries a number, and a note with the same number sits in the margin, naming the mandate clause, the limit and the fund's figure.

Above the memorandum sit a fit of 72 out of 100, a verdict to pursue at up to 5.50x instead of the amount asked, with 10 of 12 tests passing, and a strip of the twelve Fund IV tests, two of them red.

2

The banker's leverage sentence turns red

Page 5 says the $285 million unitranche is "total leverage of 5.3x Adjusted EBITDA of $54.2 million". The note beside it gives the mandate maximum of 5.75x and says the fund passed three similar deals. The seller's figure counts $7.3 million of add-backs not yet earned, and on the fund's $46.9 million it is 6.08x. At 5.75x the fund could lend $270 million.

With past pass reasons switched on, the note also links to three industrial services deals the fund passed above the same limit. Two of them are now in trouble.

3

One page, every figure with its page

The one-page screen shows each agent's step in order. The intake reader confirms the confidentiality agreement was signed on October 5. The financials extractor pulls 64 figures from the CIM, the QoE and the model, each with its page. The earnings quality analyst sorts the five add-backs; the mandate and portfolio tester runs twelve tests.

Four red flags follow, each linked to the page it came from: leverage above the limit, $7.3 million of add-backs not yet earned, a top customer at 18.4% of revenue whose contract renews on March 31, 2027, and receivables slowing from 58 to 71 days.

4

From the seller's $54.2 million to the fund's $46.9 million

The EBITDA bridge takes each add-back in turn. The ERP system, founders' pay and the Toledo branch closure are accepted, $5.1 million in all, each with a QoE page. The $3.3 million of acquisition synergies (identified, not acted on) and $4.0 million of run-rate pricing (outside the twelve months) are rejected under the fund's policy.

The leverage meter puts the seller at 5.26x, the proposal at 5.50x and the fund's figure at 6.08x. Cash interest at SOFR plus 5.25% is $25.9 million, and cover is 1.81x.

5

Daniel recommends, Elena approves, the questions go out

The screen writer sets out three options: pursue at up to 5.50x the fund's EBITDA, a $258 million unitranche at SOFR plus 5.50% with an original issue discount of 98.0 and a $90 million hold, inside every limit; pursue as asked, which needs an exception from the investment committee; or pass, with a note to the banker. Daniel keeps the first and signs at 10:04 a.m.: "I have read this screen and recommend it".

Elena Marsh approves at 10:12 a.m. Only then does the note with six questions go to Pellham Stroud, and the decision and its reason are kept.

The 6 minutes 12 seconds is the sample run's time on Lantern's 192 pages. The solution's design target is a first-pass screen in minutes for every memorandum, read the day it arrives; measure it on your own deals.

Earlier passes, what-ifs on the memo, the ranked pipeline, the twelve tests and the sponsor's record in the product

In shortOne memorandum is the story. These screens show the rest: the fund's earlier passes, two what-ifs tried on the memo itself, the ranked pipeline, the twelve tests with their sources, the sponsor's record and comparable deals, the decision and the signature, the reasons the fund passes, and the mandate every deal is read against.

Who may pass a deal, who approves pursuit, and the limits on each screening agent

In shortEach agent has one job and limits it does not cross. None of them sends anything to a banker, waives a limit, signs an indication or decides pass or pursue. The underwriting VP may pass a deal alone with a written reason; pursuing needs the head of underwriting; an indication above $100 million goes to the investment committee. Every figure cites its page, and every treatment a person changes keeps its reason.

What each screening agent reads in a CIM, and what it never opens

The vice president recommends; the head of underwriting approves pursuit

Signatures record the name, the decision, what it means and the time. Agent steps and human decisions are kept apart on the activity record; mandate changes are versioned; every pass and pursue is kept with its reason, who decided and how long it took.

Daniel and Elena, who screen and approve at the Chicago fund

From CIM and data room to a ranked pipeline, a cited one-page screen and a decision log with reasons

In shortIt reads what bankers and sponsors send, and the fund's own mandate, policy, portfolio and deal history. It returns a ranked pipeline, a one-page screen with every figure cited, the fund's EBITDA and leverage, red flags and first questions, and a decision log with reasons. The fund's systems of record stay where they are.

Deal inbox, data rooms and the pipeline CRM a screen draws on

For each, check whether it has an API, a file feed, or only screens, and who owns it.

Your screening template kept, your mandate turned into settings

No CIM figure without its page, no question sent without a decision

What a slow or loose first screen costs a direct lending fund

In shortStart from your own costs. Nobody publishes hours per screen or how often a fund pursues a deal that a full read would have stopped, so the numbers that matter are yours. A few published figures frame them: how far sellers' EBITDA overshoots, how private credit loans are performing, and what an analyst hour costs.

How many CIMs you read in full, and how long each takes

The loan a turn too high, and the deal you never read

Hours per CIM and diligence spent on deals later dropped: the return from screening faster

In shortPut in your own numbers. The calculator starts from a published analyst wage and asks you for the rest: how many memoranda you screen, the hours each takes now and after, how many deals you drop after diligence has started, what that diligence costs, and what the tool and your testing cost a year. Every default marked illustrative is an assumption to replace.

Every CIM read in full, and diligence spent only inside your limits

Underwriting hours and diligence spend per memorandum screened

A fund that receives 30 memoranda a week and reads most of them in full clears the first line easily. A fund that sees five a month will not cover a $150,000 yearly cost on time alone; its case rests on the add-back and limit checks, so test those first.

A written mandate, an add-back policy and a deal history: your underwriting team's readiness, scored

In shortPick the answer closest to your team today. Eight questions score 0 to 2; the last is for reflection. The meter adds up your score as you go, and each question says why it matters and what a strong answer sounds like.

Underwriting readiness bands, from writing the mandate to implementing

Sixteen vendor questions on add-backs, your mandate, confidential deal data and a test on your own deals

In shortAsk every vendor the same questions, us included. The answers show what the tool reads, whose rules it applies, how it treats add-backs, what evidence you keep, how it protects each deal's confidential information, and how you would measure it on your own deals.

Re-screening deals your fund has already decided, then live ones alongside the team, over ten weeks

In shortMeasure first, then test on decided deals with errors you planted, then run in parallel on live deals, then decide with four numbers. None of these steps commits you to a vendor, and your people decide every deal throughout.

1. Time five screens by step, and count last quarter's memoranda received, read in full, pursued and closed. That is your baseline.
2. Write down the rules: the mandate, the add-back policy with each test's base, the rate and shock for cover, the approval route.
3. Gather 40 decided deals, passes and pursues, with your screens and reasons. Copy ten and plant known errors: unearned synergies, a stale rate, a sector over its cap.
4. Run in parallel on live memoranda for four weeks, with your underwriters deciding every deal.
5. Decide with four numbers: agreement with your add-back treatments by type, planted errors caught, false alarms on clean deals, and underwriter minutes per memorandum.

Add-back agreement by type and planted errors caught in CIMs

CIM reviews, EBITDA add-backs, 2026 leverage and the BDC 70% test, answered for direct lenders

In shortThe questions direct lenders and buyers search for most, each answered on its own.